Link URL

1// URL helpers shared by build-description and its tests.
2//
3// normU: the key a link is looked up by in a --linkmap. Only parts that are equivalent by the URL
4// rules are normalised: scheme and host are case-insensitive (lowercased), a fragment never reaches
5// the server (dropped), a trailing slash on the path and percent-encoding of the path (decoded, so a
6// Hebrew slug compares equal to its encoded form) are folded, and per-visit tracking parameters are
7// removed. The path and the query VALUES keep their case: /Product/A and /product/a can be two
8// different pages, and the two are different keys.
9export const TRACKING = /^(utm_|visitorid$|fbclid$|gclid$|_ga$|mc_)/i;
10
11export const stripTracking = (u) => {
12  try { const x = new URL(u); let changed = false; for (const k of [...x.searchParams.keys()]) if (TRACKING.test(k)) { x.searchParams.delete(k); changed = true; } return changed ? x.href : u; } catch { return
13};
14
15const dec = (s) => { try { return decodeURIComponent(s); } catch { return s; } };
16
17export const normU = (u) => {
18  const s = String(u);
19  try {
20    const x = new URL(stripTracking(s.replace(/#.*$/, '')));
21    const path = dec(x.pathname).replace(/\/$/, '');
22    const q = [...x.searchParams].map(([k, v]) => `${dec(k)}=${dec(v)}`).join('&');
23    return `${x.protocol}//${x.host.toLowerCase()}${path}${q ? '?' + q : ''}`;
24  } catch {
25    return dec(s.replace(/#.*$/, '')).replace(/\/$/, ''); // not an absolute URL (mailto:, tel:, a relative path): compared as written
26  }
27};